Introduction

This chapter introduces machine learning concepts that support robotic perception and decision-making. You will start by contrasting supervised and unsupervised learning, and see why safety-critical robotics leans on labeled, supervised data for classification and regression tasks. From there you will see how a camera image breaks down into a grid of pixels, compare the RGB and HSV color spaces, and use color thresholding to isolate a target object into a binary mask. You will then learn how a bounding box condenses those matching pixels into a compact position and size estimate, letting a robot judge an object’s distance and steer to keep it centered in frame. The chapter closes with a hands-on decision tree classifier challenge that ties feature-based classification back to a concrete sorting task. These ideas prepare you to integrate learned components into larger robotic systems.